Impact Story: From Grad Student to AI Research Startup Cofounder

Kathleen Finlinson (Spring 2024 Impact Accelerator Program alum) shares her journey from being a graduate student to becoming an AI research startup cofounder.

I have been interested in Effective Altruism (EA) since 2012. In 2014, I founded EA Boulder when I moved to Boulder, Colorado (USA) for grad school. At this point, I already had a master's in math, and was entering a PhD program in applied math. I was already sold on AI safety as an important cause and tried to move my PhD studies in the AI safety direction, to varying degrees of success—I ended up studying complex dynamical systems and modeling biological neural nets. I also worked at some ML startups on the side while I was in grad school.

In 2018, I was offered a research job at Open Philanthropy to study AI forecasting, and I left my PhD early to take this role. There, I was studying neural net scaling laws and computation costs for training large models. Then, in 2019, I went to another ML startup founded by a friend that was using statistical modeling to detect lead pipes in cities' drinking water systems.

Meanwhile, I started a small executive coaching business mainly for EA nonprofit founders. I picked up my first client in 2020 (who I'm still working with!), and things grew slowly from there. In mid-2021, I decided to take some time off from my career to focus on meditation. I thought this period might last for a few months, but it ended up lasting 3 years. I did many meditation retreats (over 20), a lot of home meditation practice, and lived in a zen monastery for ~6 months. I continued coaching a handful of clients during this time, mainly people working in animal welfare.

In 2024, I decided to leave the monastery and come back to working in AI safety. It was a bit difficult to get my career momentum going again after such a long break, and the Impact Accelerator Program was really helpful throughout this process. It provided me with structure, energy, and companionship during my job search, as well as helping me understand what my skills are, and what problems I wanted to work on. 

By mid-2024, I was consulting with AI policymakers, advising them on AI forecasting and scenario planning. I considered spinning up a new org to do more of this and met some people who thought I was doing good work and were interested in funding it. But, largely due to the huge amount of time I'd spent networking all year, I was offered a cofounding role with Rob Long at Eleos AI Research, which studies AI sentience and welfare. Eleos released a big report in October 2024 that got some good media coverage (e.g., here and here).